Skip to main content

Table 2 Performance of rrBLUP (average r 2) on the traits of four real data sets under the traditional encoding vs. the hybrid encodings vs. the target-based encoding. We also show the improvements of the hybrid encodings over the traditional encoding

From: Does encoding matter? A novel view on the quantitative genetic trait prediction problem

Data set Traditional encoding Hybrid one (improvement) Hybrid two (improvement) Target-based
Rice: Pericarp.color 0.433 0.499 (16 %) 0.504 (16.4 %) 0.493
Rice: Protein.content 0.176 0.176 (1 %) 0.177 (1 %) 0.177
Pig: Trait 2 0.237 0.238 (1 %) 0.239 (1 %) 0.236
Pig: Trait 4 0.203 0.218 (7 %) 0.218 (7 %) 0.207
QTLMAS: Trait 1 0.358 0.36 (1 %) 0.361 (1 %) 0.36
QTLMAS: Trait 2 0.187 0.179 (-4 %) 0.18 (-4 %) 0.178
Maize: Flint 1 TASS 0.47 0.492 (5 %) 0.492 (5 %) 0.475
Maize: Flint 2 DMC 0.301 0.311 (2.5 %) 0.308 (2.3 %) 0.289
Maize: Flint 3 DM_Yield 0.057 0.07 (20 %) 0.068 (19 %) 0.062
Maize: Dent 1 Tass 0.59 0.615 (4.4 %) 0.616 (4.4 %) 0.593
Maize: Dent 2 DMC 0.562 0.58 (3.2 %) 0.58 (3.2 %) 0.582
Maize: Dent 3 DM_Yield 0.321 0.343 (8.6 %) 0.349 (8.7 %) 0.346
  1. The bold numbers are the ones with the best performance